Abstract

VLSI communication networks are wire limited. The cost of a network is not a function of the number of switches required, but rather a function of the wiring density required to construct the network. This paper analyzes communication networks of varying dimension under the assumption of constant wire bisection. Expressions for the latency, average case throughput, and hot-spot throughput of k-ary n- cube networks with constant bisection are derived that agree closely with experimental measurements. It is shown that low-dimensional networks (e.g., tori) have lower latency and higher hot-spot throughput than high-dimensional networks (e.g., binary n-cubes) with the same bisection width. Keywords Communication networks, interconnection networks, concurrent computing, message-passing multiprocessors, parallel processing, VLSI. 1 Introduction The critical component of a concurrent computer is its communication network. Many algorithms are communication rather than processing limited. Fi...

Citations

929	Computer Networks - Tanenbaum - 1981
566	Deadlock-Free Message Routing in Multiprocessor Interconnection Networks - Dally, Seitz - 1987
430	Sorting Networks and Their Applications - Batcher - 1968
362	Introduction to VLSI Systems - Mead, Conway - 1980
314	Virtual Cut-Through: A New Computer Communication Switching Technique - Kermani, Kleinrock - 1979
236	The torus routing chip - Dally, Seitz - 1986
200	hot spot’ contention and combining in multistage interconnection networks - Pfister, Norton - 1985
149	Parallel Processing with the Perfect Shuffle - Stone - 1971
132	Generalized Hypercube and Hyperbus Structures for a Computer Network - Bhuyan, Agrawal - 1984
122	Mathematical Theory of Connecting Networks and Telephone Traffic - Benes - 1965
92	A Complexity Theory for VLSI - Thompson - 1980
81	A VLSI Architecture for Concurrent Data Structures - Dally - 1987
55	The Indirect binary n-cube microprocessor array - Pease - 1977
38	The Message-Driven Processor - Dally, Fiske, et al. - 1992
38	Concurrent VLSI Architectures - Seitz - 1984
33	Wire-Efficient VLSI Multiprocessor Communication Networks - Dally - 1987
28	Design of a self-timed vlsi multicomputer communication controller - Dally, Song - 1987
21	Fat Trees: Universal Networks for Hardware-Efficient Supercomputing - Leiserson - 1985
18	The Tree Machine: A Highly Concurrent Computing Environment - Browning - 1980
18	Interconnection networks for SIMD machines - Siegel - 1979
14	The flip network in STARAN - Batcher - 1976
12	A large scale homogeneous machine - Sullivan, Brashkow - 1977
8	The Flip Network - Batcher - 1976
8	Virtual Cut-Through: A New Communication Switching Technique - Kermani, Kleinrock - 1979
4	The Hypercube Communications Chip - Seitz - 1985
3	Alignment and access of data in an array processor - Lawrie - 1975
2	Ametek 2010 product announcement - Corporation - 1987
2	A Complexity Theory of VLSI” Dep - Thompson - 1980
2	The Flip Network of Staran - Batchef - 1976
1	Single chip computers, The new VLSI building block - Sequin - 1979
1	Display File 5182:DF:85 - Technol - 1985
1	Ametek Corporation, Ametek 2010 product announcement - Ill - 1987
1	A VLSIArchitecture for Concurrenl Data Structures - Daily - 1987
1	et al., "Architecture of a message-driven processor - Daily - 1987
1	Design of a self-timed VLSI tnulticomputer communication controller - Dally, Song - 1987
1	A large scale homogeneous ma chine - Sullivan, Bashkow - 1977
1	A complexity theory of VLSI," Dep Corn put. Sci., Carnegie-Mellon Univ., Tech. Rep CMU-CS-80-140 - Thompson - 1980